| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <e5469460-209b-44f2-8b3a-1e67539e281b@arm.com>
Date: Thu, 16 May 2024 15:56:26 +0200
From: James Clark <james.clark@....com>
To: Mao Jinlong <quic_jinlmao@...cinc.com>
Cc: coresight@...ts.linaro.org, linux-arm-kernel@...ts.infradead.org,
devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-arm-msm@...r.kernel.org, Tingwei Zhang <quic_tingweiz@...cinc.com>,
Yuanfang Zhang <quic_yuanfang@...cinc.com>,
Tao Zhang <quic_taozha@...cinc.com>, songchai <quic_songchai@...cinc.com>,
Suzuki K Poulose <suzuki.poulose@....com>, Mike Leach
<mike.leach@...aro.org>, Rob Herring <robh@...nel.org>,
Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley
<conor+dt@...nel.org>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>
Subject: Re: [PATCH v1 2/3] coresight: Add reserve trace id support
On 16/05/2024 15:23, James Clark wrote:
>
>
> On 16/05/2024 04:56, Mao Jinlong wrote:
>> Dynamic trace id was introduced in coresight subsystem so trace id is
>> allocated dynamically. However, some hardware ATB source has static trace
>> id and it cannot be changed via software programming. Reserve trace id
>> for this kind of hardware source.
>>
>> Signed-off-by: Mao Jinlong <quic_jinlmao@...cinc.com>
>> ---
>> .../hwtracing/coresight/coresight-platform.c | 26 +++++++++++++++++++
>> .../hwtracing/coresight/coresight-trace-id.c | 24 +++++++++++++++++
>> .../hwtracing/coresight/coresight-trace-id.h | 11 ++++++++
>> include/linux/coresight.h | 1 +
>> 4 files changed, 62 insertions(+)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-platform.c b/drivers/hwtracing/coresight/coresight-platform.c
>> index 9d550f5697fa..d3e22a2608df 100644
>> --- a/drivers/hwtracing/coresight/coresight-platform.c
>> +++ b/drivers/hwtracing/coresight/coresight-platform.c
>> @@ -183,6 +183,17 @@ static int of_coresight_get_cpu(struct device *dev)
>> return cpu;
>> }
>>
>> +/*
>> + * of_coresight_get_trace_id: Get the atid of a source device.
>> + *
>> + * Returns 0 on success.
>> + */
>> +static int of_coresight_get_trace_id(struct device *dev, u32 *id)
>> +{
>> +
>> + return of_property_read_u32(dev->of_node, "trace-id", id);
>> +}
>> +
>> /*
>> * of_coresight_parse_endpoint : Parse the given output endpoint @ep
>> * and fill the connection information in @pdata->out_conns
>> @@ -315,6 +326,12 @@ static inline int of_coresight_get_cpu(struct device *dev)
>> {
>> return -ENODEV;
>> }
>> +
>> +static int of_coresight_get_trace_id(struct device *dev, u32 *id)
>> +{
>> + return -ENODEV;
>> +}
>> +
>> #endif
>>
>> #ifdef CONFIG_ACPI
>> @@ -794,6 +811,15 @@ int coresight_get_cpu(struct device *dev)
>> }
>> EXPORT_SYMBOL_GPL(coresight_get_cpu);
>>
>> +int coresight_get_trace_id(struct device *dev, u32 *id)
>> +{
>> + if (!is_of_node(dev->fwnode))
>> + return -EINVAL;
>> +
>> + return of_coresight_get_trace_id(dev, id);
Can we somehow make this function name distinct from the trace ID
functions. It's a bit hard to read that it's called
coresight_get_trace_id() but it doesn't actually get an ID from the
existing trace ID stuff.
>> +}
>> +EXPORT_SYMBOL_GPL(coresight_get_trace_id);
>> +
>> struct coresight_platform_data *
>> coresight_get_platform_data(struct device *dev)
>> {
>> diff --git a/drivers/hwtracing/coresight/coresight-trace-id.c b/drivers/hwtracing/coresight/coresight-trace-id.c
>> index af5b4ef59cea..536a34e9de6f 100644
>> --- a/drivers/hwtracing/coresight/coresight-trace-id.c
>> +++ b/drivers/hwtracing/coresight/coresight-trace-id.c
>> @@ -110,6 +110,24 @@ static int coresight_trace_id_alloc_new_id(struct coresight_trace_id_map *id_map
>> return id;
>> }
>>
>> +static int coresight_trace_id_set(int id, struct coresight_trace_id_map *id_map)
>> +{
>> + unsigned long flags;
>> +
>> + spin_lock_irqsave(&id_map_lock, flags);
>> +
>> + if (WARN(!IS_VALID_CS_TRACE_ID(id), "Invalid Trace ID %d\n", id))
>> + return -EINVAL;
>> + if (WARN(test_bit(id, id_map->used_ids), "ID is already used: %d\n", id))
>> + return -EINVAL;
>
> Do these returns not skip unlocking the spinlock?
>
> It might be slightly fewer changes if we update the existing
> coresight_trace_id_alloc_new_id() to add a new "only_preferred" option.
>
> Then use the existing system id allocator which already handles the lock
> and unlock properly:
>
> static int coresight_trace_id_map_get_system_id(struct
> coresight_trace_id_map *id_map, int id,
>
> bool only_preferred)
> {
> ...
> spin_lock_irqsave(&id_map_lock, flags);
> /* prefer odd IDs for system components to avoid legacy CPU IDS
> id = coresight_trace_id_alloc_new_id(id_map, id, true,
> only_preferred);
> spin_unlock_irqrestore(&id_map_lock, flags);
> ...
>
> I suppose the end result is the same as your implementation, but it
> trades making one existing function slightly more complicated instead of
> adding some new ones.
It's also not that obvious that there is the new reserve function, but
you still free the ID with the same coresight_trace_id_put_system_id().
Another benefit of adding arguments to the existing functions is that we
keep just ...get...() and ...put...(). 'Reserve' implies some other new
mechanism, but it's really a normal get. I think we should do one of
these two options for the top level API:
#1 (when id != 0, then it's an "only preferred" preferred ID:
coresight_trace_id_get_system_id(int id)
coresight_trace_id_put_system_id(int id)
#2
coresight_trace_id_get_system_id()
coresight_trace_id_get_system_id_resrv(int id)
coresight_trace_id_put_system_id(int id)
Powered by blists - more mailing lists